SLAC-PUB-7949 CERN-TH 98-297 SU-ITP-98 54

Stabilization of Sub-Millimeter Dimensions: The New Guise of the Hierarchy Problem

Nima Arkani-Hameda, Savas Dimopoulosb and John March-Russellc

hep-th/9809124 v2 7 Oct 1998

SLAC, Stanford University, Stanford CA 94309, USA Physics Department, Stanford University, Stanford CA 94305, USA c Theory Division, CERN, CH-1211, Geneva 23, Switzerland

Abstract
A new framework for solving the hierarchy problem was recently proposed which does not rely on low energy supersymmetry or technicolor. The fundamental Planck mass is at a TeV and the observed weakness of gravity at long distances is due the existence of new sub-millimeter spatial dimensions. In this picture the standard model elds are localized to a 3 + 1-dimensional wall or 3-brane". The hierarchy problem becomes isomorphic to the problem of the largeness of the extra dimensions. This is in turn inextricably linked to the cosmological constant problem, suggesting the possibility of a common solution. The radii of the extra dimensions must be prevented from both expanding to too great a size, and collapsing to the fundamental Planck length TeV,1. In this paper we propose a number of mechanisms addressing this question. We argue  that a positive bulk cosmological constant  can stabilize the internal manifold against  expansion, and that the value of  is not unstable to radiative corrections provided that the supersymmetries of string theory are broken by dynamics on our 3-brane. We further argue that the extra dimensions can be stabilized against collapse in a phenomenologically successful way by either of two methods: 1 Large, topologically conserved quantum numbers associated with higher-form bulk U1 gauge elds, such as the naturally occurring Ramond-Ramond gauge elds, or the winding number of bulk scalar elds. 2 The brane-lattice-crystallization of a large number of 3-branes in the bulk. These mechanisms are consistent with theoretical, laboratory, and cosmological considerations such as the absence of large time variations in Newton's constant during and after primordial nucleosynthesis, and millimeter-scale tests of gravity.

1 New Guise of the Hierarchy Problem

A new proposal for solving the hierarchy problem was recently introduced 1, 2, 3 which circumvents the need for supersymmetry or technicolor. Instead the hierarchy problem for the standard model SM is solved by bringing the fundamental Planck scale down to the TeV scale. Gravity becomes comparable in strength to the other interactions at this scale, and the observed weakness of gravity at long distances is then explained by the presence of n new large" spatial dimensions. Gauss' Law relates the Planck scales of the 4 + n-dimensional theory, M, and the long-distance 4-dimensional theory, Mpl,
2 n Mpl  rn Mn+2

1 2

where rn is the size of the extra dimensions. Putting M  1 TeV then yields
n rn  10,17+ 30 cm

For n = 1, r1  1013 cm, so this case is excluded since it would modify Newtonian gravity at solar-system distances. Already for n = 2, however, r2  1 mm, which happens to be the distance where our present experimental knowledge of gravitational strength forces ends. For larger n, 1=rn slowly approaches the fundamental Planck scale M. While the gravitational force has not been measured beneath a millimeter, the success of the SM up to  100 GeV implies that the SM elds can not feel these extra large dimensions; that is, they must be stuck on a 3-dimensional wall, or 3-brane", in the higher dimensional space. Thus, in this framework the universe is 4 + n-dimensional with fundamental Planck scale near the weak scale, with n  2 new sub-mm sized dimensions where gravity, and perhaps other elds, can freely propagate, but where the SM particles are localised on a 3-brane in the higher-dimensional space. The most attractive possibility for localizing the SM elds to the brane is to employ the D-branes that naturally occur in type I or type II string theory 4, 2 . Gauge and other degrees of freedom are naturally con ned to such D-branes 4 , and furthermore this approach has the obvious advantage of being formulated within a consistent theory of gravity. However, from a practical point of view, the most important question is whether this framework is experimentally excluded. This was the subject of 3 where laboratory, astrophysical, and cosmological constraints were studied and found not to exclude these ideas. 1

There are also a number of other important papers discussing related suggestions. Refs. 5 examine the idea of lowering the GUT scale by utilizing higher dimensions. Further papers concern themselves with the construction of string models with extra dimensions larger than the string scale 6, 7, 8 , and gauge coupling uni cation in higher dimensions without lowering the uni cation scale 9 . There are also two important papers by Sundrum. The rst deals with the e ective theory of the low energy degrees of freedom in realizations of our world as a brane 10 , while the second is concerned with the topic of radius stabilization 11 , and with which our analysis has much in common. In our framework the hierarchy problem becomes the problem of explaining the size and stability of the large extra dimensions. The main purpose of this paper is to exhibit mechanisms which accomplish these objectives, and examine some aspects of their phenomenology. Since a rather wide collection of possible stabilization mechanisms are discussed in this paper, only some of which we believe to be successful, we think it useful to provide the reader with a guide to our main results: In Section 1.1 we discuss a very general consistency constraint on the bulk cosmological constant; and in Section 2 we describe some basic kinematics pertaining to the radial oscillation eld, whose mass will turn out to provide signi cant constraints on stabilization scenarios. In particular this is the constraint that will sometimes force us to have a large conserved integer parameter in our models. In Section 3 we show that the properties and limits on such light radial oscillation elds can be discussed in a way that is independent of the details of the precise radius-stabilization mechanism. We also brie y describe the reasons for the cosmological safety of this scenario. The most important results of this paper are contained in Section 4 where we discuss long-distance IR and, particularly, shortdistance UV stabilization mechanisms, and put these together to obtain a variety of complete stabilization models. We nd that two methods of UV stabilization are particularly attractive: brane-lattice-crystallization" discussed in Section 4.2; and topological stabilization" discussed in Section 4.3. Finally in Section 5 we present a summary of our results.

1.1 The Hierarchy and the Bulk Cosmological Constant.

Let us begin with some necessary conditions that must be satis ed to ensure the existence of large radii. As we know from experience with our 4-dimensional world, to ensure that our three ordinary spatial dimensions are very large the radius of curvature of the universe must be no less than the present horizon size. This leads to the require2

ment that the cosmological constant of the universe is less than the critical density. An identical line of reasoning for the case of n-extra dimensions also leads to an upper limit on the bulk cosmological constant as we now explain 12 . The curvature radius Lcurv of the bulk space in the presence of energy density or an  e ective cosmological constant, , in the bulk, is  n+2 !1=2  : 3 Lcurv  M  This curvature radius must be larger than the physical size of the transverse dimensions rn in order to insure that the bulk space does not split o " into separate in ating universes separated by horizons of size Lcurv, or collapse into black holes. This gives  an upper bound on  12 :  !4=n   M4+n M  4 M
pl

This constraint will play an important role in what follows. It already implies that  the magnitude  must be smaller than the fundamental scale of M. This was to be expected since in this case there is one scale in the problem and the bulk would split into a collection of non-communicating 1= TeV size regions, outside of each others' particle horizons. An important corollary of this is that one cannot use the Scherk-Schwarz mechanism to break supersymmetry at M since this would induce a bulk cosmological constant of the order of M4+n , which exceeds the limit Eq. 4. Of course the e ective 4-dimensional cosmological constant measured at long distances greater than the size of the extra dimensions must to a very high degree of accuracy vanish. This can be achieved by cancelling the wall and bulk contributions against on another:  0 = f 4 + rn n 5 We see that if the bulk energy is negative, a positive f 4 will cancel the 4-dimensional cosmological constant, while if the bulk energy is positive, we need a negative f 4. Clearly a positive f 4 is reasonable; if the wall can uctuate in the extra dimensions, f 4 is just the tension of the wall, and provides the correct sign kinetic term for the Nambu-Goldstones of spontaneously broken 4 + n-dimensional Poincare invariance which live on the wall. This reasoning seems to exclude the possibility of a negative f 4, since this gives the wrong sign kinetic term to the Nambu-Goldstones. This is however only a problem if the Nambu-Goldstone elds are indeed present, that is, if the 4 + n-dimensional Poincare invariance is spontaneously broken. On the other hand, suppose that the wall is stuck" and cannot uctuate in the extra dimensions, 3

due to explicit breaking of 4 + n-dimensional Poincare invariance. As an example, we can consider twisted sector elds living at an orbifold xed point. In the language of string theory the wall could be an orientifold rather than a D-brane. In this case, f 4 is just the wall energy density acting as a source for gravity, but there are no NambuGoldstones on the wall to receive a wrong-sign kinetic term. Another way of saying 4 this is as follows. The wall can have an energy density as a source for gravity fgrav, 4 4 and a tension under bending" fbend. It is fgrav which should appear in Eq.5. If the 4 + n-dimensional Poincare invariance is only spontaneously broken, its non-linear 4 4 realization forces fgrav = fbend, as they both come from expanding the term in the action Z p 6 , d4 ,gind f 4; 4 4 where gind is the induced metric on the wall. Since fbend 0, we have fgrav 0. On the other hand, if the 4 + n-dimensional Poincare invariance is explicitly broken, there need not be any relationship between the two. Indeed, if the wall can not uctuate, 4 4 e ectively fbend = 1, while fgrav can be nite and of any sign. We will therefore allow the possibility that a brane can make a net negative contribution to the 4-dimensional cosmological constant, which provides us with the freedom to consider stabilization mechanisms that give either positive or negative bulk energy densities.1 Given Eq. 4 we learn that if our wall is the only brane, then its e ective walllocalized cosmological constant, f 4, is bounded above by  Mpl n,2=2n f  M M 7  This is not too severe a constraint though, varying between 10 TeV for n = 2, to  108 GeV for n = 6. Of course, the relation, Eq. 5, can be turned around to determine  the e ective bulk cosmological constant, , given f . A natural assumption for the walllocalized cosmological constant, given our state of knowledge of the standard model interactions on the wall, is f 4 = 1 TeV4  M4. Thus in this case  !2 M  8  = M4+n M 
pl

is the value of the bulk cosmological constant necessary to cancel the total long-distance cosmological constant in our world. Note that this value is indeed always less than the upper bound Eq. 4 arising from the bulk curvature constraint.
1

We thank Eva Silverstein for discussions about this point.

Later in Section 4.2 we will consider stabilization mechanisms that utilize many branes populating the bulk. In this case the bounds Eqs. 4 and 7 are modi ed by the total brane number Nwall. A di erence with the previous case is that the curvature radius must now only be greater than the inter-brane separation rn =Nwall1=n. We are assuming the best case situation of equally spaced branes which leads to the weakest bound. The reason for this is that the branes themselves are localized sources of curvature of the opposite sign, so that at long distances compared to the inter-brane separation, the curvature of the bulk averages out to zero. From this follows the generalized curvature constraint  !4=n 4+n M  : 9   NwallM M
pl

The IR cancellation of the e ective cosmological constant in 4-dimensions is expressed by  0 = Nwallf 4 + rnn : 10 Imposing this leads to the following bound on the wall-localized cosmological constant 1  Mpl 2n,4=n : 4 M4 f   n,2=n M 11  Nwall The cosmological constant is bounded from below from another consideration. As we will see later in Section 3, there are light gravitationally coupled particles in the  spectrum whose mass2 is proportional to  see Eq. 36. The requirement that these particles do not con ict with measurements of gravity imply that they weigh  more than a meV and consequently put a lower limit on . This in turn implies that the large size of the new dimensions in most, but not all cases studied here, cannot be  solely due to the smallness of . Additional dynamics to boost the size of the extra dimensions are necessary. This can easily come about if there is a conserved charge in the system, analogous to baryon number. Just as humans are large because they carry large baryon number, the extra dimensions can be large because they carry some large charge Q. In some of our examples, this charge corresponds to a large number of walls Q  Nwall 1. In others, it is a topological charge k. Note however that in some special cases, it is not necessary to use a large conserved charge. For example, as we discuss in Section 4.3 if the fundamental scale M is pushed to  10 TeV while the wall contribution to the cosmological constant f 4 is kept at  1 TeV4, then topological charges k  1 are adequate. This is not too unnatural a situation, especially considering that a loop factor could easily supply such a suppression to f 4. 5

1.2 Stable and Calculable Hierarchy

In this paper we will not search for dynamical mechanisms where the hierarchy between the size of the extra dimensions and the fundamental scale is calculable. We will instead be content to enforce this hierarchy by choosing the bulk cosmological constant to be small and or the above-mentioned topological or other charge to be large. This is analogous to the early days of the minimal supersymmetric standard model MSSM 13 where the soft supersymmetry breaking terms were postulated without any reference to a dynamical mechanism which generates them. The idea there was that since the problem of supersymmetry breaking is connected with the cosmological constant problem it seemed premature to adopt a speci c SUSY-breaking mechanism and it seemed more prudent to study consequences that were independent of the details of the SUSY-breaking mechanics. Similarly, in our new framework the hierarchy and cosmological constant problems are even more closely intertwined so we will adopt a similar philosophy of not insisting on a detailed dynamical mechanism for a calculable  hierarchy and will be content to instead parametrize our ignorance by a choice of  and an integer Q Nwall or k. The second aspect of the hierarchy problem is its stability against radiative corrections. In the MSSM this is guaranteed by low energy supersymmetry, which protects the Higgs mass against large radiative corrections. Presumably, the analogous ques tion in our framework is the behaviour of the pair of parameters ; Q under radiative corrections. The integer Q is automatically protected since it refers to charge of a  con guration. Since  is a bulk cosmological constant one can imagine two possibilities. One is that whatever solves the cosmological constant problem will also prevent   from becoming as large as the cuto M. The second more explicit and perhaps  more satisfactory viewpoint is to invoke bulk-supersymmetry to protect  from large radiative corrections. Indeed, as pointed out in reference 2 , if supersymmetry is broken solely on our 3-brane by an amount  M  1 TeV, the Fermi-Bose splittings that this induces in the bulk are miniscule  TeV2=Mpl  10,3 eV and therefore the bulk  cosmological constant  is protected by the approximate bulk-supersymmetry. It should be emphasized that stabilizing large dimensions is inherently easier than stabilizing Planck-scale dimensions. In the latter case, quantum gravitational e ects are necessarily important and can not be ignored. However, precisely because we are interested in large radii, the details of short distance physics are largely irrelevant and a classical or semi-classical analysis su ces. We will consider this point more explicitly in Section 4. 6

2 Kinematics of Radius Stabilization

Suppose that we have an N -brane embedded in a space with N large spatial dimensions and n small dimensions we wish to stabilize. The total action is comprised of a bulk part,  Z q n+N ,1 R +  , L 1+N +n x , det G 12 Sbulk = , d matter + : : : ; 1+N +n M and a brane part,

Sbrane = ,

d1+N x

induced , det g1+N 

 N +1 + : : : ; f

13

where Lmatter is the Lagrangian of bulk gauge or scalar elds, and the ellipses denote higher-derivative terms that can be ignored in the regime of interest as we will demonstrate below. Take the background metric for the 1 + N + n-dimensional spacetime to be of the form 0 1 1 C; g = B ,Rt2gIJ 14 @ A 2g ,rt ij where R is the scale factor of the N -dimensional space, and r is the scale factor of the internal n-dimensional space, with geometry set by gij where detgij  = 1. With this metric the Ricci scalar is  _!  _! R + N N , 1 R 2 +2n r + nn , 1  r 2 +2Nn r_R + nn , 1 ; 15 _ ,R = 2N R R r r rR r2 where the internal curvature term is present for n-spheres  = 1, but vanishes for tori  = 0, and we have ignored a similar curvature term for the large dimensions. After integrating over all spatial coordinates we obtain, Z _ S = dtLKE R; r_ , RN Vtotr; 16 where the total potential is given by

Vtotr = Vbulk + Vwall Vwall = f N +1 Vbulk = rn , nn , 1Mn+N ,1 rn,2 + Vmatterr
where,

17 18

Z Vmatterr = , dn x rn Lmatter:

7

After integrating the R and r terms by parts, the kinetic part of the action for the radii, R and r, becomes 0  _ !2  _ !1 Z _ R + nn , 1  r 2 + 2Nn r_R A : 19 S = ,MN +n,1 dt RN rn @N N , 1 R r rR Note the overall negative sign of these kinetic terms. This is connected to the wellknown phenomenon that the conformal mode of gravity has the opposite sign kinetic term to the transverse graviton kinetic term and which bedevils attempts at de ning quantum gravity via the Euclidean functional integral. _ _ In any case there is clearly an extremum of the action with R = r = 0, when the condition @RRN VtotrjR=R0;r=r0 = 0, and similar with @R ! @r  is met. These imply for R0 6= 0

Vtotr0 = 0; 0 Vtotr0 = 0:

and

20

This is as one would have naively expected. However, because of the negative sign for the kinetic term for the radial degrees of freedom, the stability analysis for such static solutions has to be treated with care. The analysis starts by expanding the action, Eq. 19, in small uctuations around the extremum: Rt = R0 + Rt, and rt = r0 + rt. Then to quadratic order, and de ning
 R=R0 and  r=r0, the expansion gives the coupled equations of motion  ! !  ! ! N N , 1 Nn
= 0 0
; 21 2 Nn nn , 1 0 ! where 00 2 00  0 2Vtot 22 !2 = 1 MrN+n,1 rr0n = 1 r0 Vtot1r0 : 2  2 MN , 0 N +1 Here MN +1 is to be understood as the e ective Planck mass in the large N + 1dimensional spacetime M4  Mpl. We now search for oscillating solutions, 
;  = expi t
0; 0 of the stability equations. From Eq. 21, 2 is thus given by the eigenvalues of the matrix  ! !2 0 ,Nn + nN N + n , 1 0 N N , 1 ; 23 namely,
2

= 0, and

 = nNN , 1 1 !2: +n, 8

24

The zero eigenvalue just corresponds to the fact that R0 is a at direction since, by assumption, there is no potential for R. The crucial expression is Eq. 24, which gives us the condition for stability of our static solution. Stability requires 2 0, which for N 1 implies 00 !2 0  Vtotr0 0: 25 This is the main result of this Section. Even though it seems trivial that stability is equivalent to requiring the second derivative of the potential around the extremum to be positive, this condition is a priori not at all obvious given the negative kinetic terms for the radii elds. As an example of this consider the case N = 0, which corresponds to r being thought of as the radius of a Friedman-Robertson-Walker universe. In this 00 case stability requires ! 2 0, or equivalently Vtotr0  0. This accords with our usual understanding: for example take the only term in V to be a positive cosmological term Vtotr = rn. Then around the minimum at r = 0 the solution is unstable to in ationary growth as we expect. The end result of this analysis is simply that we can think in terms of a total potential V r that one can minimize to nd the stable static solutions for the size of the internal dimensions. Also note that from Eqs. 24 and 22 we can extract the mass of the canonically normalized radial oscillation eld we will refer to as the radion" in the case of interest, N = 3, n arbitrary: 2 V 00 26 m2 = nn1+ 2 r0 Mtotr0 radial 2 pl Notice that as a consequence, the magnitude of is related to the deviation r from the equilibrium radius r = r0 + r via r : 27 r0 Mpl

3 Model-independent limits on light radions

Before we move on to the very important issue of the explicit nature of possible radius stabilization mechanisms as discussed in Section 4, it will be useful for us to examine some model-independent features of all these mechanisms. These include the existence of a light radial oscillation eld , with known couplings to standard model elds. Although such a eld seems to be dangerous, we will argue below that it satis es the various limits. 9

To see that independent of the details of the stabilizing potential there is an upper bound on the mass of the radial excitation eld it is useful to consider a general form for the bulk stabilizing potential Vbulkr. Around the equilibrium position this potential can be well approximated by the sum of just two powers of r:2
 Vbulkr = M4 Axa + Bxb : 28 Here we have introduced the dimensionless radius variable x  rM. In particular, following on from the discussion in Section 2, for a stable minimum we study potentials of the form  N Vbulkr = M4 x + x ; ; 0; 29 or
 Vbulkr = M4 x , x ; 0: 30 As we discuss in Section 4 the dimensionless parameter is a measure of the size of the e ective bulk cosmological constant, and acts to prevent the radius from expanding to in nity. In contrast, the N or  terms prevent collapse to the UV, and arise from either inter-brane interactions, or from the kinetic energy of topologically quantized bulk gauge or scalar elds. As we will soon see, to get a large radius requires a small , and or a large N or . Requiring the cancellation of the e ective 4-dimensional cosmological constant at the minimum of these potentials leads to the equations
0 Vbulkr0 = 0; Vbulkr0 + Nwallf4 = 0:

31

Here we have allowed for the possibility that there is more than one wall or brane in the bulk, Nwall  1. These provide localized sources of curvature in principle of either sign as discussed in Section 1.1. However, for simplicity, we have assumed that all the 4 branes have broadly similar such energy densities fgrav;i ' f4, and of the same sign. More general possibilities can also be analyzed. In any case, the equations 31 can be used to determine r0 and the required value of Nwallf4 in terms of the basic model-dependent parameters of the potential, ; ; , etc. Alternately, we can nd the values of these parameters necessary to produce a
In this paper we will not explicitly consider potentials of the form Vbulk r  r f logr , for some function f , with only a single such term dominantly contributing to the potential energy near the equilibrium position. See Ref. 3 for a discussion of such potentials.
2

10

the stabilizing parameters and N are determined to be: 9 = , + x10 = given f4 0: N = , + x0 ; In the case of the potential Eq. 30, = ,
1 x0 ;
0

desired internal radius x0 = r0M in string units. De ning the useful dimensionless combination f4 32  Nwall M 4


33

9 = given f4 0: 1 ; x ;

34

Now, by equipartition, the second derivative of the general potential V rbulk of Eqs. 29 and 30 around the minimum is given by V 00  V r0bulk=r02. In addition the mean bulk value of the cosmological constant is de ned by    Vrr0n : 35 0 Thus using the de nition of the canonically normalized radial excitation, Eq. 26, it is easy to see that physical mass of the radial excitations is  36 m2  M n : radial 2+   But now we can apply the curvature radius bound on , Eq. 4, to nd  !4=n 2 2 M mradial  M M Nwall  Nrwall 37 2
pl

independent up to the O1 coe cients we have dropped of any details of the stabilizing potential or mechanism. Evaluating this for the most conservative case of Nwall = 1 and for the desired values of M leads to a mass for the radial eld that varies between 10,2 eV or less for n = 2, to  20 MeV or less for n = 6. Note that the reason why the  radion mass is much smaller than M is that  must be relatively small to allow large extra dimensions. So, in all the models for radius stabilization that we consider, the radion eld will , be very light with m2  rn 2, at most  20 MeV for n = 6. Thus it is necessary to

11

study the model-independent limits on such light radions to make sure that the entire scenario is not excluded. To do this, we have to determine the coupling of radions to SM elds on our wall. At rst, it seems that there is no direct coupling of the radion eld to SM elds. The reason is that the couplings of SM elds to gravity all come from the induced metric on our wall, which, if the possible Nambu-Goldstones on the wall are turned o so the wall is at in the extra dimensions, depends on g but not the radion elds gmn . However, this argument is not correct and the radion eld does indeed couple to SM elds as we now show.3 Let us go to the e ective theory at distances large compared to the size of the extra dimensions. The e ective action is    ! Z q 4 x ,detg 2 1 + n +


R + g  @ @ , m2 2 + L ; g  ; d  ,Mpl   SM  Mpl 38 where  are the SM elds. Notice that since the e ective 4-dimensional Planck scale depends on the size of the extra dimensions, there is dependence in the coe cient of R, and that there is no explicit dependence on in the SM part of the Lagrangian as expected. However, there is kinetic mixing between ordinary gravity and the eld, speci cally if we expand around a at metric g =  + h , there is a mixing of the form @ 2h . Thus, even though there is no direct coupling of the SM elds to , one is  induced through this mixing. This can be seen more clearly if we rst perform a Weyl rescaling to remove the dependence in front of the usual graviton kinetic term. The coupling to SM elds then comes from the scale-invariance violating part of the SM lagrangian; the leading interaction is

LSM , = M TSM :  pl

39

Note that interactions are suppressed for relativistic particles, while it has comparable strength to gravity for non-relativistic particles. Suppose that is massless. As far as the long-range force between non-relativistic particles is concerned, this just amounts to a rede nition of Newton's constant GN ! GN;non,rel:, while the Newton constant governing the interaction of gravity with light GN;rel: retains the standard value GN;rel: = GN . However, the successful predictions for the gravitational de ection of light as well as Big-Bang Nucleosynthesis assume GN;non,rel: = GN;rel: at least to within a few percent. Moreover, since the long range force between non-relativistic masses has been measured down to  1 mm without revealing any deviation from Newtonian gravity, the mass of must be pushed up above  1mm,1  10,3 eV. This model3

We thank Raman Sundrum and Riccardo Rattazzi for setting us straight on this point.

12

independent constraint, which was also discussed in 11 , is the most important limit on light radions. We now move on to consider other possible limits on light radions coming from cosmology and astrophysics. There are two classes of worries. The rst is that not only the radion but all of its Kaluza-Klein excitations can be produced in the early universe and in stars, leading to such well-known problems as the over-e cient cooling of supernovae. This concern is identical to the problem of bulk graviton overproduction which was studied in 3 , and found to in some cases namely n = 2 extra dimensions to constrain but not rule out our scenario. The second concerns oscillations of the radion eld itself around its minimum. These may overclose the universe, and further, since these oscillations correspond to changing the size of the extra dimensions, they also lead to an oscillating 4-dimensional Newton's constant, which can be problematic.4 Therefore we now brie y discuss some aspects of the cosmology of radion elds. We will adopt here the same attitude taken in 3 . There, limits were put on the highest temperature T up to which the universe could be considered normal", that is, with the extra dimensions stabilized and energy density dominated by the radiation on our wall. Since on-the-wall interactions can produce gravitons which escape into the bulk and which in turn can variously a ect the expansion rate of the universe during nucleosynthesis, overclose the universe, and unacceptably distort the background photon spectrum when they decay, the normalcy temperature T was limited to  few MeV to  1 GeV for n = 2 , 6. Fortunately, T  1 MeV in all cases with n = 2 marginal, so that the successful predictions of primordial nucleosynthesis can still hold in our scenario. In our present analysis of radion cosmology, we will be content to show the cosmological safety of the scenario at temperatures  T. Namely, we will assume that the early universe at temperatures  T evolved into a state with the radion stabilized, with the energy density stored in radion oscillations small enough to never overclose the universe. We will show that this is enough to guarantee negligible variations in GN , and that subsequent interactions with SM elds on the wall will not signi cantly excite the radion away from its minimum.5 First, note that since T TeV, the Hubble expansion rate at all times of interest satis es H  T 2=Mpl 1mm,1 , so that the expansionary friction" term can not
We remark that high-frequency oscillations of GN of su ciently small amplitude around a mean value equal to the standard value of GN can be accommodated, despite the fact that in such a case dGN =dt can be signi cantly larger than the usually quoted bounds. 5 A full discussion of the very early universe cosmology in our scenario, in particular the worry that an early period of in ation could lead to a form of the Polonyi problem involving will appear in Ref. 16 .
4

13

stop from oscillating. Further, since is so light and gravitationally coupled, it is essentially stable cosmologically. The energy density stored in its oscillations redshifts away as 1=R3 , so that =T 3 is invariant. It is easy to see that in order for to never dominate the energy density of the universe, we must have
crit: now ,9 40 3 T 3  Tnow  3  10 GeV: Using  m2 2, this is enough to show that the variations in GN are miniscule at all epochs T  T: p GN = n r  ,12 41 G r M  m M  10 :

Furthermore, interactions with the SM elds can not signi cantly excite the radion into oscillation. Note that it is only the excitations of and not its associated KK modes which would correspond to changing the radius of the extra dimensions and hence varying GN  on cosmological scales. This single mode has couplings suppressed by the ordinary 4-dimensional Planck scale Mpl, and it is therefore very di cult to excite. Quantitatively, the rate at which collisions of SM particles dump energy into is T7 42 _  M2 : The total amount of energy dumped into during a Hubble time is then T5 M ; pl leading to an unobservably small variation in GN q  !1=2 GN  T5  10,18 : G m M  1 TeV4M
N
pl pl pl

pl

pl

43

44

Of course, this is hardly surprising. Recall that at temperatures below T, the energy dumped into the bulk gravitons and together with all of their KK excitations never overclose the universe. Even if all of this energy was somehow transferred into moving the single mode , we already found that as long as the energy did not overclose the universe, the variations in GN are negligible. Similar comments apply to radion excitation in stars: the energy lost to the production of together with all its KK excitations are safe for the same reason as bulk graviton production is safe see Ref. 3 , while the single mode is too weakly coupled to be perturbed enough for signi cant variations of GN to be observable. For instance, 14

in the collapse of SN1987A over a time tSN  1s, the variation in GN can be estimated as above, yielding  n+7 ! GN = n r  TSN tSN 1=2 1; 45 G r M n+2 m2M 2
N SN
0

pl

even in the worst case TSN  100 MeV, n = 2. Thus the local variation of GN is harmless for any number of extra dimensions n = 2 , 6.

4 Radius Stabilization Mechanisms

We now turn to some explicit mechanisms by which internal dimensions may be stabilized at a radius much greater than the fundamental Planck length  M,1. Two issues must be distinguished in discussing radius stabilization: the mechanism by which the internal dimensions are prevented from collapsing to 1=M , and the mechanism by which they are prevented from expanding to a size much larger than a millimeter or fermi.

4.1 Generalities
The most obvious idea for limiting the expansion of the internal dimensions is to employ a component of the potential energy that scales like the volume of the internal space: V  rn . Such an e ective potential energy density results from a positive bulk cosmological constant , which gives V r  rn as shown in Section 2. As we have already discussed the size of this bulk cosmological constant must be small. However, while we have no compelling explanation for the size of this bulk cosmological constant, it is interesting that its smallness can at least be stable under radiative corrections. Suppose that the short-distance theory of gravity perhaps string theory is supersymmetric, with the supersymmetries broken only on the walls at a scale  M  jfj  1 TeV. It is easy to see that the Bose-Fermi splittings induced in the bulk supergravity multiplet are then 2

M jm2 , m2 jbulk  Nwall M  bose fermi 2

pl

46

so that the quantum corrections to the bulk cosmological constant would be of order
 jquantj  jm2 , m2 jbulk 4+n=2 : 47 bose fermi 15

The ratio of the quantum correction to the tree value is bounded above once the curvature constraint is used, and we nd  !2n+4=n M quant 1: 48  Mpl tree Therefore, the small value of the cosmological constant can be technically natural. We now turn to the ways in which the radii of the extra dimensions can be stopped from collapsing to small values. We will see that a wide range of mechanisms are in principle possible, leading to a variety of power-law potentials of the form 1=r` for various `. One minimal possibility is if the compact manifold has positive curvature, in which case Vbulkr  rn , Mn+2 rn,2 49 As can readily be seen from Eq. 34, this will require a large positive value for the ratio = Nwallf4=M4 , which can arise if we have a con guration of a large number Nwall 1 of branes. This possible brane lattice crystallization", together with various generalisations, will be discussed in the next subsection. Alternately, if we wish to compactify on manifolds with no curvature tori, the ultraviolet stabilization can be provided by dynamics conserving a topological number k, which we will explore in Section 4.3.

4.2 Radius Stabilization from Brane Lattice Crystallization

The largeness of the internal dimensions compared to 1 TeV,1 can arise from the existence of a large conserved number of branes populating the bulk. There can exist inter-brane forces which act like the Van der Walls and hard-core forces between atoms in a crystal. The inter-brane distance is set by these forces, and might be quite small, but the size of the whole internal space is set by the total number of branes, just as the total extent of a crystal is set by the number of atoms, rather than just the inter-atom distance which is much smaller.

I Minimal scenario.

We will motivate stabilizing the extra dimensions with large brane numbers by considering the minimal example of compact manifolds with positive curvature, which together with a positive bulk cosmological constant give a bulk potential of the form Vbulk = rn , Mn+2 rn,2 50 16

For n = 2, the curvature contribution to the potential is constant and does not play any role in radius stabilization although it does contribute an extra term to the e ective 4-d cosmological constant. For n 2, however, this potential has a stable minimum. From Eq. 34, we nd that a large value for is needed,  Mr0n,2. For simplicity, we will assume that all the branes are broadly similar with f4  M4. Then, we must have a large number of branes  2n,2=n Nwall  Mr0n,2  Mpl : 51 M Numerically this varies from Nwall  1010 for n = 3 to Nwall  1020 for n = 6. Although this is a large number it is not so large as to lead to problems. Specifically, we note that there is a constraint on the total number of branes that can populate the internal dimensions. If the transverse inter-brane separation becomes comparable to 1=M , then there will be new light open string modes that arise from strings starting on one brane and ending on a neighbor. Thus the maximum number of branes that can occupy the extra dimensions is  2 n M n  Mpl  1032 ; 52 Nwall; max  r0  M which is considerably greater than the necessary number Eq. 51. However, with such a large number of branes, it is obviously important to ensure that some dynamics forces them to spread out in the bulk and not sit on top of each other. This can easily be arranged. We know that there is a gravitational force between the branes, and if they carry any sort of like sign gauge charge there will also be an opposite gauge force between them with exactly the same dependence on inter-brane separation. In fact, when the charge density is equal to the tension T , there can be an exact cancellation of the inter-brane forces. This is what happens in the case of supersymmetric D-branes. Polchinski's now classic calculation of the forces between D-branes demonstrated that the forces due to Ramond-Ramond gauge elds precisely cancelled the gravitational forces in the supersymmetric limit, as they must for a pair of BPS states, which satisfy T = . If there is a mismatch between the charge and tension of the branes, the net force between a pair of branes can be made repulsive, forcing them to spread out uniformly in the bulk. Of course, we must now take the inter-brane potential energy into account in the energetics, but interestingly, this e ect is parametrically of the same order as the terms in the potential we already 17

have. By Gauss' law, the potential between branes falls o with the inter-brane separation r according to the coulomb potential in the transverse n dimensions Vint:r  2=rn,2 . If we rst imagine just two 3-branes populating the internal space the potential energy varies as V r  M4 rM1n,2 : 53 Here we have taken the e ective net charge density on the wall to be M4 , as we would expect if supersymmetry is broken at a scale jf j  M. The interbrane distance can be estimated from balancing this repulsive force against a  bulk cosmological constant term V r  rn . Imposing the cancellation of the 4-dimensional cosmological constant Eq. 5, leads to an inter-brane separation rI M4 54 rI n,2  M n  1 TeV2,n: +2
 

What happens when Nwall branes occupy the internal space? One may think that the size of the internal volume will just be Nwall times larger than rI n calculated above. However this is incorrect. The reasons for this are two-fold. The rst is that, unlike in a normal crystal, there is no necessity that the inter-brane forces are screened. Thus the total potential energy density due to the inter-brane 2 forces increases as Nwall, just as the gravitational potential in a star, and the UV stabilizing part of the potential has the form 2 V  M4Nwall rM1n,2 ; 55  where r is now roughly the total extent of the system. The second reason why the two brane calculation is modi ed is that the equation for the cancellation of the e ective 4-dimensional IR cosmological constant is modi ed to Eq. 10. Putting all parts of the potential together, we have 2  V rtot  M4Nwall rM1n,2 + rn + Nwallf 4: 56  Solving for the size of the system gives,  2 !1=2n,1 r0  Nwall 6 : 57  Mn, 18

 However  can be eliminated by imposing Eq. 10, with jf j  M leading to

n+4    Mn,2 ; 2= Nwall and thus the nal expression for r0

58

r0  Nwall M

1=n,2

59

Utilizing the formula for the required size of the extra dimensions, r0nMn+2 = 2 Mpl, we can solve for the necessary brane-number  2n,2=n ; 60 Nwall  Mpl M exactly the expression Eq. 51.  Notice that if one substitutes this value back into the equation for , Eq. 58, then one nds  !4=n   Mn+4 M  ; 61 M

of the hierarchy problem in this framework is the bulk cosmological constant problem. As discussed above, there is one other requirement that needs to be satis ed. The mean curvature radius on scales smaller than the inter-brane separation needs to be larger than the inter-brane separation itself. The average inter-brane transverse separation is now  n !1=n r0 1  Mpl 4=n2 ; rI  N M M 62 wall   whilst the curvature radius resulting from our potential is 1  Mpl 2=n : Lcurv  M M 63   For the case of n 2 where the above analysis applies, one always has Lcurv rI as required. In addition if the supersymmetries of string theory are broken only by on-the-wall dynamics at a scale  M  1 TeV, then the mass splittings so induced among 19

pl n+4 , showing that indeed one component which is smaller than the naive value M

the bulk supergravity multiplet p  M2=Mpl Nwall, and a bulk cosmological are 2 =Mpl Nwall 4+n arises. Then the ratio of this new constant of order   M  term to the  term is  ! quant:  M 2n+4=n 1: 64 tree Mpl Therefore the value of the bulk cosmological constant can still be technically natural in the case of a large number, Nwall, of branes. In summary, we have made a number of simplifying assumptions which can be questioned and modi ed. These include the simpli cation that all 3-branes are broadly similar and have tensions and charge densities  1 TeV4 with a mismatch that is also of order 1 TeV4. Nevertheless, it is encouraging that the large-brane-number scenario for stabilizing the volume of the internal dimensions at large values passes the rst tests.

II Non-extensive Bulk Cosmological Constant.

There is another interesting possibility where the size of the extra dimensions is completely explained by a large brane number without needing to invoke another small parameter the small bulk cosmological constant in the above analysis. Suppose that the the IR potential is ra , for a n with the normal rn term being sub-dominant. The bulk potential then reads N2 65 V rbulk  M n,walln,2 + ra  6r With this potential we still need the same large brane number Nwall  Mr0n,2. The size of  is to be  !2a+2,n=n 4+a M   M M : 66 For a = n , 2, the required value for  agrees with the natural value  M4+a. This is intriguing, since a = n , 2 is precisely the power associated with curvature terms! Of course, a compact manifold with positive curvature makes the wrong-sign contribution to the potential, but we can choose the compact manifold to have negative curvature. For instance, genus g 1 Riemann surfaces have negative Euler characteristic and hence negative average curvature by the Gauss-Bonnet theorem. In order to stabilize more than two dimensions in this way, we can compactify on direct products of such Riemann surfaces, which will then give the correct exponent and the correct sign in the potential of Eq.65. 20
pl

III Casimir forces between branes.6

Another potentially attractive idea for UV stabilization at the quantum level is to use the Casimir force to maintain the size of the internal space 14 . The e ective 4d potential energy density corresponding to the Casimir e ect in a 4 + n-dimensional spacetime is C V r  r4 ; 67 where C is a calculable coe cient in any given model. Even with a general nonextensive stabilizing potential,
V  ra this leads to a inter-brane distance of  1=4+a rI  C : 68  Given that the natural" value of  is expected to be M4+a, this clearly doesn't allow us to stabilize at large radii. What about many branes? The problem is that, when we go to Nwall 3-branes in the bulk, the Casimir energy does not increase with Nwall for n  2. But the total wall cosmological constant Nwallf4 does, and thus the situation gets worse. In summary, the Casimir force idea, even with a large brane number Nwall 1, fails to stabilize the internal dimensions at large radii, at least under the simplifying assumptions we have made.

4.3 Topological Stabilization

One of the most attractive ways of preventing collapse is to imagine that there is a topologically conserved quantity which holds up the size of the extra dimensions. A prototypical example of this is provided by the monopole stabilization mechanisms discussed in Ref. 15 and in the context of our scheme by Sundrum 11 . Consider the simple case of two extra dimensions and where the internal manifold has the topology of a 2-sphere, S 2. Further suppose that in the bulk there exists not only the graviton, but also a U1 gauge eld, which might naturally be a Ramond-Ramond RR gauge eld of the string theory in question. Then it is possible to take the gauge eld con guration on S 2 to be topologically non-trivial with quantized monopole number" k the rst
6

This possibility was also analyzed by Sundrum 11 .

21

Chern number of the U1 bundle given by7 1 Z H 2 = k: 69 2 S2 If the area of the S 2 is denoted V2 then we have H  k=V2 and since the kinetic term for the U1 gauge eld is expressed in form notation, with M 4 denoting 4-dimensional Minkowski space M2 Z SKE  g2 4 2 H ^ H; 70 M S we have that the 4-dimensional potential energy density of the monopole eld on the S 2 scales like 2 2 V  M2 Vk : 71 g In other words we get an energy density that scales like k2M2=g2r2. For large enough monopole number, k, this will stabilize the internal S 2 at any desired size. This basic mechanism has a wide variety of generalizations. One such is to use the topological invariants of the higher-form RR gauge elds that naturally arise in the type II and type I string theories with D-branes. Let us recall here that for stabilizing n 2 dimensions topologically, we must work with compact manifolds of zero curvature i.e. tori, since otherwise the curvature term will dominate the dynamics and we must revert to the analysis of the previous subsection.
2

I Higher-form RR elds

Denote the manifold of the extra n dimensions by E n, and suppose that the bulk theory contains an n , 1-form U1 gauge eld, with n-form eld strength F n. Then once again there is the topological invariant Mn,2 Z H n = k: 72 2 En The kinetic energy of H n is the generalization of the usual 1-form gauge kinetic term Mn Z SKE;n  g2 4 n H n ^ H n; 73 M E and thus the potential energy density depends on the volume Vn of E n as 2 1 V  g2M n,4 Vk : 74 n  22

We will always use H for eld strengths of gauge elds that live in the bulk. Quite often we will think of these as being RR gauge elds.
7

In the case of the chiral type IIB string theory there exists 1 and 3-form RR eld strengths and a self dual 5-form RR eld strength together, of course, with their magnetic duals. There also exists the usual NS-NS 3-form eld strength. The type I string theory has a 3-form RR eld strength and it's 7-form magnetic dual. Thus using the invariants so far described, it is natural to stabilize 1, 3, and 5-manifolds. However, invariants that lead to 1=rn potentials for E n are not the only possibility. Consider the situation in which our 3-brane world is the boundary of a set of higher-dimensional branes which are in turn embedded in the full 4 + ndimensional space. We can then use topological invariants of the world-volume gauge elds of these higher-dimensional branes to stabilize the internal dimensions. To make this clear consider the following very simple example: In the n = 4 case take the internal manifold to be E 4 = T12  T22. Suppose further that there exist 2 5-branes that intersect at the position of our 3-brane but are perpendicular in the extra 4 dimensions, so that one 5-brane lives in M 4  T12 and the second lives in M 4  T22. Then we have the two topological invariants 1 Z F =k; i = 1; 2; 75 2 Ti2 i i where Fi, i = 1; 2 are world-volume U1 2-form eld strengths of the rst and second 5-brane. The brane-localized kinetic terms for these gauge elds then leads to an e ective 4-dimensional potential energy density of the form 2 2 2 2
V r  Mr2k1 + Mr2k2 ; 76 1 2 where r1 and r2 are the radii of the two T 2's. Note that since we have used tori, there is no negative curvature term  ,M2r2 in the potential. This, then, is an UV stabilizing potential for E 4 = T12  T22 not of the form 1=r4 . Clearly this type of mechanism admits many generalizations. Finally, one can also consider higher reducible" invariants such as the second Chern class of the usual 2-form U1 eld strength de ned wrt a 4-manifold 1 Z H 2 ^ H 2 = c ; 77 2 82 E4 but such invariants typically lead to a potential energy varying as r with  0.

II Metric topological invariants.

23

Purely metric topological invariants are possible, for example the Euler number of a 2-manifold component E 2 of the internal space Z = 21 E2 R; 78  where R is the curvature 2-form. Other possibilities include the Pontrjagin classes of the tangent bundle of the internal manifold. However, because the leading term in the gravitational e ective action is only linear in the curvature, this does not provide a UV stabilizing potential unless higher derivative terms, such as Z
S  Mn trR2  79 are included in the e ective action. For the simple case of n = 2 this leads to a potential V  2M2=r2 . For this to balance, at the appropriate r0, even a bestcase stabilizing potential of the form M5 r, an Euler number of  Mpl=M 3=2 is required. So clearly the internal manifold is very highly curved. In particular, R the leading gravitational action M4 R dominates the other terms by an amount Mpl=M1=2, and leads to an unacceptably large bulk cosmological constant. This seems to be a generic problem with this type of topological stabilization, although we have not investigated the question in detail.

III Scalar- eld and other non-gauge invariants.8

One can also imagine stabilizing the size of the internal space by the use of nongauge or metric topological invariants. For example, consider a complex scalar eld that lives on a 1-dimensional higher brane that has as boundary our 3brane. Then the phase of this eld can wind as an S 1 cycle of the internal space is transversed, with topologically conserved winding number Z k= 1d : 80 Once again the kinetic energy of this con guration increases as the size of the internal space is reduced, and thus a stabilizing potential results. In order to stabilize more than one dimension in this way, we can have n di erent scalar elds living on n di erent 4-branes which have the 3 dimensions of our 3-brane in common but have mutually orthogonal fourth spatial dimensions. The i'th scalar eld can wrap around the 4th dimension of the i'th 4-brane, generating a potential , , of the form r1 1 +


rn 1. Together with a bulk cosmological constant giving a
S

Gia Dvali has independently considered this possibility. We thank him for discussions.

24

potential  r1


rn , this can stabilize all the n dimensions. More sophisticated scalar eld invariants are also conceivable, the Hopf winding number of the map  : S 3 ! S 2 being one among many such examples. In general scalar eld invariants lead to quite similar results to the gauge eld topological stabilization mechanisms, but possibly without the natural advantage of gauge elds of their constrained couplings. For instance it is easy to arrange that the stabilizing gauge elds do not lead to dangerous avor-changing neutral current processes on the wall, while this requires additional input in the scalar case.

IV Phenomenologically successful topological stabilization.

In the previous subsections we have seen that a variety of UV stabilizing potential energy densities of the general form M4k2=rM   are possible. We now wish to get an idea of the numerical values the various parameters must take to stabilize the radii at the desired sizes. From Eq. 33, we have that  Mpl 2 =n  1 TeV 4 4 2  jNwall f j M r  81 k  0  M M4 M :  where the inequality comes from imposing the constraint that the radion is heavier than  1mm,1. In the above examples of topological radius stabilization the quantity k is directly proportional to the monopole" number. From this expression the smallest k clearly occurs when the ratio =n is as small as possible. As an example, if = 1 and there 6 extra dimensions, then for M  1 TeV, k  3  102 is required to stabilize at a su ciently large radius. If = 2 and there are 6 extra dimensions, then k  105 is necessary. The = 2 case is particularly interesting since it is the rst case we can realize with gauge- eld topological invariants rather than scalar eld invariants. The worst case, requiring the largest k, occurs when =n takes on its largest value. A typical worst-case" is provided by the irreducible topological stabilization mechanism involving bulk RR elds for example. This gives = n, and leads to k  1015. Note that in the special case with = 1 and n = 6, if we are willing to move the fundamental scale M up to  10 TeV, while keeping Nwallf4  1 TeV4, we can get away with k  1. This may not be unnatural, after all, one could easily imagine that the scale f is  10 smaller than the scale M due to partial cancellations up to some 1 or 2-loop order. Also recall that as shown in Section 4.1, the small value of the bulk cosmological constant is unexplained but is at least technically natural, if SUSY is primordially broken on the walls. 25

In summary we have shown that the topological stabilization mechanism successfully meets all our phenomenological requirements, with a price of a large, but in some cases not too large integer k.

V Corrections to leading-order potentials.

Finally, one may worry that in the regime of interest, when r  r0, the semiclassical reasoning that we have applied to the leading-order kinetic and non-derivative terms in the e ective action su ers from large corrections due to the presence of other terms. Such corrections are, in actual fact, entirely negligible. For example, if one included higher-order derivative terms, such as Z
S  Mn,4 4 n H p ^ H p ^ H p ^ H p; 82
M E

negligible unless k  1030 . Such statements generally apply for r  r0, and are basically due the fact that r0 1 TeV,1. This is not quite trivial because of the potentially large dimensionless factor k which could have overcome this suppression. In any case we see that the leading-order analysis is entirely su cient unless we are interested in physics at radii r r0.

in the e ective action, then they would lead to corrections in the 4-dimensional e ective potential energy density, V , of order 4 83
V  M4 M kr 3n  0 at the minimum r0. Compared to the leading kinetic term this is a fractional change of order  !
V  k2 M 4 ; 84 V M
pl

5 Remarks and Summary

The hierarchy problem in our framework is replaced by the problem of obtaining large new dimensions, of a size which varies between a millimeter and a fermi depending of the number of new dimensions, in a theory with a much smaller fundamental length  TeV,1. In this paper we exhibited mechanisms which provide such large extra dimensions. These mechanisms relied on two ingredients: 26

A large conserved integer Q, which can be a large number Nwall of branes, or the topological charge k of the vacuum con guration. This large integer should be regarded as analogous to the net conserved baryon number which accounts for the large size of macroscopic objects relative to that of atoms. The necessity for such a large number was not forced on us by the need for large internal dimensions, but rather by the requirement that the radial oscillation eld or radion" be su ciently heavy to have escaped tests of gravity at the millimeter-scale and above. The value of Q depends on the details of the stabilization scenario; it varies from Nwall  1010 to Nwall  1020 in the brane-lattice-crystallization scenario, while in the topological stabilization scenario it varies from k  1 to k  1015. A small bulk cosmological constant, analogous to the 4-dimensional cosmological constant whose smallness accounts for the size of our universe relative to the Planck length. However, as we discuss in detail in, for example, Section 4.1, the value of this bulk cosmological constant is stable against radiative corrections if supersymmetry-breaking of order the fundamental Planck mass  M  1 TeV takes place on the 3-branes. Of course we must still impose a ne tuning to get a vanishing e ective 4-dimensional, brane-localized cosmological constant in the IR in our world. This is expressed in Eq. 5 or 10, depending on the stabilization scenario. A valid criticism of our analysis is that we have not provided a dynamical framework in which, for instance, the largeness of Q or k is explained. As discussed in the introduction our viewpoint on this issue is that this is closely analogous to the situation in the MSSM where soft supersymmetry-breaking operators of order 1 TeV are introduced 13 . With the advent of many quantum- eld-theoretic QFT models of dynamical supersymmetry breaking it is commonly believed that the problem of the size of these soft operators has been solved, at least in principle. However, from a fundamental vantage-point this belief is not correct. Concretely, what is the situation in the standard model or MSSM, where the usual reduced Planck mass Mpl  2  1018 GeV is taken as fundamental? We must now explain the ratio of this Planck scale to the weak scale  1015. There too the dilaton runaway problem" prevents us from having a calculational framework for this number. This point is important to emphasize. Although in the context of QFT dynamical SUSY breaking solves the hierarchy problem, in that it generates the small scale by dimensional transmutation, in the context of string theory the couplings and thus the scale of SUSY breaking are dynamical, and 27

there is a ground state at zero coupling with unbroken supersymmetry 17 . This means that there exists no known solution to the hierarchy problem in usual 4-dimensional QFT once it is embedded in string theory. Therefore both frameworks face similar challenges.

Acknowledgements
It is a pleasure to thank Philip Argyres, Gia Dvali, Michael Graesser, Vadim Kaplunovsky, Yaron Oz, Riccardo Rattatzi, Eva Silverstein and Raman Sundrum for valuable discussions. SD thanks the CERN theory group for its hospitality. The work of NAH is supported by the Department of Energy under contract DE-AC03-76SF00515. The work of SD is supported in part by NSF grant PHY-9219345-004. The work of JMR is supported in part by an A.P. Sloan Foundation Fellowship.

References
1 N. Arkani-Hamed, S. Dimopoulos and G. Dvali, hep-ph 9803315, to appear in Phys. Lett. B. 2 I. Antoniadis, N. Arkani-Hamed, S. Dimopoulos and G. Dvali, hep-ph 9804398, to appear in Phys. Lett. B. 3 N. Arkani-Hamed, S. Dimopoulos and G. Dvali, hep-ph 9807344. 4 See for example: J. Polchinski, TASI lectures on D-branes, hep-th 9611050; C. Bachas, Lectures on D-branes, hep-th 9806199. 5 K.R. Dienes, E. Dudas and T. Gherghetta, hep-ph 9803466 and hep-ph 9806292; K.R. Dienes, E. Dudas,T. Gherghetta and A. Riotto, hep-ph 9809406. 6 I. Antoniadis, Phys. Lett. B246 1990 377. 7 P. Horava and E. Witten, Nucl. Phys. B460 1996 506; E. Witten. J.D. Lykken, Phys. Rev. D54 1996 3693; E. Caceres, V.S. Kaplunovsky, I.M.Mandelberg, Nucl. Phys. B493 1997 73. 8 G. Shiu and S.H.H.Tye, hep-th 9805157. 9 C. Bachas, hep-ph 9807415. 28

10 R. Sundrum, hep-ph 9805471. 11 R. Sundrum, hep-ph 9807348. 12 P. Argyres, S. Dimopoulos and J. March-Russell, hep-th 9808138; to appear in Phys. Lett. B. 13 S. Dimopoulos and H. Georgi, Nucl. Phys. B193 1981 150. 14 P. Candelas and S. Weinberg, Nucl. Phys. B237 1984 397. 15 Y. Okada, Phys. Lett. B150 1985 103. 16 N. Arkani-Hamed, S. Dimopoulos and J. March-Russell, in preparation. 17 M. Dine and N. Seiberg, Phys. Lett. B162 1985 299.

29